Liran Bachar
Liran Bachar
@microsoft-github-policy-service agree company="Habana" From: microsoft-github-policy-service[bot] ***@***.***> Sent: Monday, 13 March 2023 18:31 To: microsoft/DeepSpeed ***@***.***> Cc: Liran Bachar ***@***.***>; Mention ***@***.***> Subject: Re: [microsoft/DeepSpeed] Fix pipeline module evaluatiion when contiguoush...
@tjruwase Can you please review?
@tjruwase Can you please approve running the workflows?
@tjruwase Thanks for reviewing this change. I have made changes to address your comments. Now there is no need to pass HALF_DTYPE as compiler define. all functions will be templated...
Added templated invoker to help selecting the implementation The map stores function pointers to templated functions, the key is the type enum. At initialization all supported dtypes are templated and...
@tjruwase not sure why amd-mi200 test failed
I never encountered compilation being aborted. I used GCC 11.4.0 and now I tested on 12.2.0 Both compilers take about 30 seconds to complete on my CPU (xeon server). 11.4.0...
I was unable to reproduce this issue. I tried the following nvcr docker images: nvcr.io/nvidia/pytorch:24.06-py3 (torch 2.4, cuda 12.5) nvcr.io/nvidia/pytorch:24.03-py3 (torch 2.3, cuda 12.4) nvcr.io/nvidia/pytorch:23.06-py3 (torch 2.1, cuda 12.1) Could...